The relationship between the ocean's normalized radar cross section (NRCS) at 14.6 GHz and the surface wind vector is derived using the 3 months of Seasat microwave scatterometer (SASS) measurements. The derivation is based on the statistics of the SASS observations, and no in situ measurements are required, other than a mean global wind speed, which comes from climatology. The frequency distribution of the global wind vectors observed by SASS is assumed to be a bivariate normal probability function. A NRCS model function is found that maps the assumed wind vector statistics into the observed SASS NRCS statistics. This function is compared with a NRCS model coming from the Joint Air Sea Interaction Experiment (JASIN) and with aircraft scatterometer measurements. The results indicate that the statistically derived NRCS model is an improvement over the JASIN model, which was based on a limited number of in situ anemometer measurements.

Improved second generation normalized radar cross section (NRCS) and brightness temperature (TB) models and associated wind retrieval algorithms are derived for the Seasat microwave radiometer SMMR and scatterometer SASS. The derivation of the NRCS model is based on the assumption of a Rayleigh distribution of wind speeds, and no in situ anemometer measurements are used. Furthermore, the NRCS model derivation is designed to preclude, as much as possible, systematic errors in the polarization and incidence angle relationships. A constant power law NRCS model is used, except for nadir observations. The nadir NRCS for winds above 15 m/s falls off faster with increasing wind speed than is predicted by a constant power law relationship. The TB model derivation consists of finding the wind‐induced emissivity coefficients, modifying the 37‐GHz atmospheric absorption coefficients and removing biases in the TB observations. The TB biases are found to be stable except for the 18‐GHz channels, which experience large, time‐dependent biases. The NRCS and TB models are incorporated into new wind retrieval algorithms, which are used to process the SASS and SMMR 3‐month data sets. Small residual systematic errors in the SASS winds (±0.5 m/s or less) are found. A histogram of the SASS winds closely resembles a Rayleigh distribution. The SASS winds are compared with 1623 National Data Buoy Office (NDBO) buoy observations, and a 1.6‐m/s rms discrepancy, with a −0.1‐m/s bias, is found. The SASS and SMMR winds are compared on a 150‐km cell‐by‐cell basis, giving 123,000 wind comparisons for the 3‐month period. The comparisons are done using eight different combinations of three SMMR channels. Good agreement is found between the SASS and SMMR winds, except for two of the channel combinations that show little, if any, skill in retrieving wind. Over the SASS primary off‐nadir swath, the SMMR and SASS wind agreement ranges from 1.3 to 2.2 m/s, depending on the channel combination. For the SMMR versus SASS nadir wind comparisons, the agreement slightly degrades. The SMMR winds appear to be more noisy than the SASS winds for winds below 3 m/s. These results indicate that the Special Sensor Microwave Imager (SSM/I), to be launched in 1986, will have the capability to measure the near‐surface wind speed to an accuracy of about 2 m/s.

This study reveals that the power‐law form of the Seasat A scatterometer system (SASS) empirical backscatter‐to‐wind model function (SASS 1), when combined with the sum‐of‐squares (SOS) wind retrieval algorithm, does not uniformly meet the instrument performance specification requirements. Analysis indicates that the horizontally polarized (Hpol) and vertically polarized (Vpol) components of the benchmark SASS 1/SOS wind retrieval system relating signal strength (backscatter) to wind speed yield self‐consistent results only for a small mid‐range of speeds at larger incidence angles and for a somewhat larger range of speeds at smaller incidence angles. An approach is presented that differs from previous calibration studies: here the internal Vpol versus Hpol consistency of the model is examined by the use of a set of pairwise collocated SASS‐produced winds, where one member of a wind pair (UVV) derives from only Vpol backscatter measurements and the other (UHH) from only Hpol measurements. This data set was created by extracting from the Seasat mission Geophysical Data Record (GDR) data base all pairs of SASS winds of the form (UVV, UHH) such that UVV and UHH are observations separated by no more than 5 min and 50 km. The set contains 377,289 such pairs. Comparisons of SASS 1/SOS derived wind data with wind data taken off the coast of Scotland during JASIN and with wind data from NDBO buoys off the U.S. Atlantic, Pacific, and Gulf coasts further underscores the shortcomings of the SASS 1/SOS wind retrieval system. These in situ wind comparisons to SASS indicate geographical differences in the retrieved scatterometer winds that are potentially attributable to environmental differences such as sea surface temperature. Taken together, the in situ comparisons to SASS, and the SASS Hpol to Vpol wind intercomparison indicate that the SASS 1/SOS wind retrieval system appears deficient in retrieving some winds, particularly from backscatter measurements made at higher incidence angles. We also find that Hpol backscatter measurements show greater sensitivity than Vpol backscatter measurements to wind speed variations above 10 m s−1, while the reverse holds for wind speeds less than 10 m s−1. It should be emphasized that it is the geophysical algorithms that lead to the errors cited here, not the instrumentation or the principles of scatterometry. In displaying these sources of error we have been obliged to outline the algorithmic methodology in a manner of presentation not previously attempted either in the published literature or in the voluminous code documentations. We hope that this material will be of value in itself to users of scatterometer data who seek to enhance their understanding of how such data are derived.

A series of data assimilation experiments is performed to assess the impact of Seasat A satellite scatterometer (SASS) wind data on Goddard Laboratory for Atmospheric Sciences (GLAS) model forecasts. The SASS data are dealiased as part of an objective analysis system utilizing a three‐pass procedure. The impact of the SASS data is evaluated with and without temperature soundings from the NOAA 4 Vertical Temperature Profile Radiometer (VTPR) instrument in order to study possible redundancy between surface wind data and upper air temperature data. In the northern hemisphere the SASS data are generally found to have a negligible effect on the forecasts. In the southern hemisphere the forecast impact from SASS data is somewhat larger and primarily beneficial in the absence of VTPR data. However, the inclusion of VTPR data effectively eliminates the positive impact over Australia and South America. This indicates that SASS data can be beneficial for numerical weather prediction in regions with large data gaps, but in the presence of satellite soundings the usefulness of SASS data is significantly reduced.

Owing to the antenna configuration, and the biharmonic nature of the anisotropic return of the probing radar, the SEASAT scatterometer wind reports consist of as many as four wind directions at each point of observation. A systematic methodology is proposed for selection of the true direction from among these, in the absence of any in situ observations. The techniques employed are those familiar to meteorological analysts. They were subjected to rigorous testing for the JASIN workshop, with two teams performing independent analyses. The agreement of the two analyses was striking. Comparison with wind directions from the JASIN data set, and those estimated from the analyzed pressure field, show agreement largely within the SEASAT specifications. It is concluded that the methodology employed is suitable for SEASAT scatterometer data users.